JP2006515187A5 - - Google Patents

Description

Apparatus and method for tracking an internal target site without using an embedded fiducial

  The present invention relates generally to an apparatus for improving the accuracy and effectiveness of a surgical treatment, and more particularly to tracking the location of a target site that is displaced during treatment by breathing and other patient movements.

  Various therapies require precise tracking of the location of the target site in order to apply or add treatment to the target site. In radiation therapy and radiosurgery, for example, a tumor can be destroyed by an ionizing radiation beam that kills cells in the tumor. Since it is desirable to direct the radiation beam only to the tumor and not to the healthy tissue surrounding the tumor, accurate aiming of the beam to the tumor is most important in radiation therapy. The goal is to minimize the exposure of surrounding healthy tissue to radiation while focusing a large amount of radiation on the tumor. In order to properly distribute the radiation dose to the tumor, the direction of the radiation beam is generally adjusted during the treatment to locate the tumor.

  State-of-the-art radiosurgery systems such as Accuray's Cyberknife System use stereo on-line X-ray images during treatment to increase the accuracy of radiation therapy. The position of a landmark (eg, a skull) such as a patient's bone can be identified with high accuracy by using a Cyberknife stereoscopic X-ray camera system. Thus, this high precision X-ray camera system can be used to treat a target site if the position of the target site relative to a landmark such as a bone remains constant. However, this X-ray camera system cannot be used to identify the position of the target site when the position of the target site relative to a landmark such as a bone changes. This is because the target, eg, a tumor, generally does not appear in the X-ray image and is not visible. For example, target sites in the patient's abdomen or chest cannot be treated with this method alone.

  Accurate beam aiming is not difficult if the tumor is in a non-moving body part such as the brain, but aiming is difficult if the tumor is in or near a moving body part such as the lungs. Tumors located near the lungs move as the patient inhales and exhales, requiring continuous adjustment of the radiation beam direction. Since changes in the location of the tumor are not necessarily related to changes in the location of the external surface of the patient, the tumor cannot be easily tracked based solely on external measurements. For example, placing an external sensor on the patient's chest and tracking the movement of the sensor does not provide accurate information about the location of the tumor within the chest cavity. The reason is that some soft tissue structures can move in the other direction when the bone moves in one direction. The location of the tumor cannot be determined by the X-ray system. This is because, in most cases, neither the target site nor the surrounding soft tissue appears in the X-ray image. The two-dimensional nature of the X-ray image impairs the accuracy with which the radiation is applied to the target site. Furthermore, even if X-rays are accurate and provide sufficient data, real-time tracking of the target site cannot be performed with X-ray images alone because the patient is exposed to excessive radiation.

  Attempts to improve target site tracking accuracy include the use of internal fiducials. For example, US Pat. No. 6,144,875 discloses a method for tracking a target site by implanting a small gold marker that appears in an x-ray image in the abdomen of a patient prior to radiation treatment. Once internal fiducials are embedded, they are imaged periodically (periodically) with a stereoscopic X-ray camera system and their location is accurately identified. Based on the position of the marker, the position of the tumor is accurately determined.

Unfortunately, the use of internal fiducials in the above manner has disadvantages. First, x-ray imaging is too slow and too invasive to track respiratory motion in real time. X-ray systems are discontinuous, allowing only to be determined at certain time intervals, eg every 10 minutes. Second, the fiducial implantation is an invasive and expensive procedure. The reason is that the procedure is usually performed under the supervision of a computer tomography (CT) device and in the presence of a surgeon. The requirement to have a surgeon not only increases the cost of the procedure for the patient, but also exposes the surgeon to ionizing radiation. Furthermore, there are real risks of complications that can result from fiducial placement.
US Pat. No. 6,144,875

  For the above reasons, it is desirable to provide an apparatus and method for accurately tracking a patient's target site without using internal fiducials, and the present invention is directed to this purpose.

  A method and apparatus are provided for locating an internal target site during therapy without using an implant fiducial. The method includes generating a plurality of first images each representing a related space including an internal target site, generating a live image of the related space during treatment, and converting the live image to one of the first images. And a step of matching with one. Since the first image shows the internal target site, the step of matching the live image to one of the first images identifies the location of the target site regardless of whether the second image itself indicates the location of the target site. . The first image may be any three-dimensional image such as a CT scan, a magnetic resonance image, and an ultrasonic image. The live image can be, for example, an X-ray image.

  The present invention can be used with one or more real-time sensors to track the location of a target site in real time. In order to achieve real-time tracking, the signal from the real-time sensor is associated with the location of the target site. A correlation model is generated by taking an X-ray image and simultaneously reading the signal from the real-time sensor and then using the X-ray image to identify the most consistent three-dimensional image that indicates the target location. Once the correlation is established, the location of the target site can be tracked in real time during the treatment by reading the signal from the real time sensor almost continuously.

The present invention is a method for locating an internal target site during therapy comprising:
Generating a plurality of first images of a patient's associated space, the associated space including the internal target site, and each of the plurality of first images indicating an internal target site;
Generating a live image of the relevant space at a predetermined time during treatment;
Determining a position of an internal target site at the predetermined time by identifying one of the plurality of first images that most closely matches a live image. In the method, the plurality of first images may include one of a computer tomography scan, an ultrasound image, and a magnetic resonance image. In the method, the plurality of first images may be three-dimensional images. In the method, generating the plurality of first images includes generating a first computer tomography scan taken at a first time point of a respiratory cycle and a second time taken at a second time point of the respiratory cycle. Generating a computer tomography scan; and generating at least one computer-generated intermediate three-dimensional image representing the associated space at a third time point between the first time point and the second time point. May be included. Generating the at least one intermediate three-dimensional image comprises interpolating between a first computer tomography scan and a second computer tomography scan; and first and second computer tomography scans; Changing one of the scans. The step of generating the at least one intermediate three-dimensional image includes the step of continuously transforming the first three-dimensional image into a second three-dimensional image obtained at a different time from the first three-dimensional image. May be included. The plurality of first images may be generated at preselected times in the patient breathing pattern. The method further includes calculating a plurality of digitally reconstructed X-ray images for each of the plurality of first images, wherein each of the plurality of digitally reconstructed X-ray images may represent an associated space from a specific viewing angle. Each of the first images may indicate a point in time of the respiratory cycle, and each digitally reconstructed X-ray image may indicate rotation and translation of the associated space. In the method, identifying one of the plurality of first images that most closely matches the second image may include comparing the digitally reconstructed x-ray image with the second image. The comparing step may include a step of comparing the digitally reconstructed X-ray image and the second image pixel by pixel. The second image may include an X-ray image. The step of generating the plurality of first images may be performed independently of the generation of the second image. The plurality of first images may be generated before treatment, and the second image may be a live image generated during treatment. The position of the internal target site may be expressed in a treatment room coordinate system. The method uses a time interval Δt _sensor Connecting the real-time sensor read in step 1 to the outer body part of the patient, reading the signal s from the real-time sensor, and Δt _sensor Generating a corresponding image of the relevant space simultaneously with a larger time interval Δt, and a plurality of first matching most closely with the corresponding image to generate a correlation between the position of the target site and the signal Identifying one of the images. Generating the correlation includes reading a signal at a plurality of different times and generating a corresponding image of the associated space to create a set of signal data and a corresponding set of image data; Adapting to a set of signal data, adapting a second curve to the corresponding set of image data, and comparing the first curve and the second curve may be included. The method also includes a time t _sensor By reading the signal s from the real-time sensor, identifying the position of the signal s in the first curve, and identifying the point y in the second curve corresponding to the signal s in the first curve, _sensor Estimating the position of the target site at. The method comprises the steps of periodically taking a new X-ray image of the relevant space during treatment and simultaneously reading the corresponding signal and updating the correlation with the new X-ray image and the corresponding signal. It may further include. The real-time sensors include infrared tracking, force sensors, anemometers, strain gauges, laser distance sensors, and various sensors based on tactile, auditory, ultrasonic, magnetic, mechanical, and visual principles. One of the following. The real-time sensor is a first sensor and may further include coupling one or more additional sensors to the patient's body to track patient movement. The method includes determining a patient's breathing pattern, and predicting a position of a target site at a later time based on the breathing pattern and the plurality of first images taken at different times of the breathing pattern. It may further include.

Another invention is an apparatus for tracking the location of an internal target site within a patient comprising
An imaging device configured to generate a live image of an associated space including the internal target site;
A storage device that stores a plurality of three-dimensional images, each of the three-dimensional images indicating a pre-generated image of a related space and an internal target site, each of which is generated before treatment;
A processor coupled to the image device and the storage device,
A processor includes a first set of computer readable instructions for selecting an optimal 3D image that most closely approximates the live image, and an internal target site at the time the live image is generated based on the optimal 3D image And a second set of computer readable instructions for determining the position of the computer. The apparatus further includes a plurality of digitally reconstructed X-ray images stored in the storage device, each of the plurality of digitally reconstructed X-ray images being associated with one of the plurality of three-dimensional images, and a processor Can use the digitally reconstructed X-ray image to select the optimal three-dimensional image. Each of the plurality of digitally reconstructed X-ray images may reflect a different displacement of the associated space due to movement of the patient's body. The plurality of three-dimensional images may include one of a computer tomography scan, a magnetic resonance image, and an ultrasound image. The plurality of three-dimensional images interpolate between the first three-dimensional image, the second three-dimensional image, and the first and second three-dimensional images and one of the first and second three-dimensional images. And at least one synthesized three-dimensional image generated by transforming the image. The imaging device may include at least one X-ray generator, and the live image may be a stereoscopic X-ray image. A detection system coupled to the processor, wherein the detection system generates a live image so that the processor generates a correlation between the motion of the outer body part and the position of the target site; At the same time, the movement of the outer body part can be read. The detection system includes infrared tracking, force sensors, anemometers, strain gauges, laser distance sensors, and various sensors based on tactile, auditory, ultrasonic, magnetic, mechanical and visual principles. One of the following. Further comprising a patient breathing pattern stored in the storage device, the processor includes a third set of computer readable instructions for generating a three-dimensional image of the relevant space at one or more preselected times of the breathing pattern. May be included. A beam generator may be further included that directs the treatment beam to the location of the target site in response to the processor command. The plurality of three-dimensional images may include at least one of a computer tomography scan, a magnetic resonance image, or an ultrasound image.

Yet another aspect of the invention is an apparatus for tracking the location of a target site within a patient,
Means for generating a plurality of first images of a relevant space inside a patient, wherein the relevant space includes a target site, and each of the plurality of first images indicates a target site;
Means for generating a live image of the relevant space at a predetermined time during treatment;
Means for determining a position of an internal target site at the predetermined time by identifying one of the plurality of first images that most closely matches a live image. In the apparatus, the means for generating the plurality of first images includes means for generating a first computer tomography scan taken at a first time point of the respiratory cycle and at a second time point of the respiratory cycle. Means for generating a second computer tomography scan to be taken and at least one computer-generated intermediate three-dimensional image representative of the relevant space at a third time point between the first time point and the second time point For generating. The means for generating the at least one intermediate three-dimensional image interpolates between a first computer tomography scan and a second computer tomography scan, and first and second computer tomography -It may include means for changing one of the scans. The apparatus further includes means for calculating a plurality of digitally reconstructed X-ray images for each of the first images, each of the plurality of digitally reconstructed X-ray images representing a relevant space from a specific viewing angle. The means for identifying one of the plurality of first images may include means for comparing the digitally reconstructed x-ray image with a live image. The device also has a time interval Δt _sensor Means for detecting movement of the outer body part of the patient, _sensor Means for reading the signal and generating a corresponding image of the relevant space at the same time with a larger time interval Δt and the corresponding image for generating a correlation between the position of the target site and the signal And means for identifying one of a plurality of first images that most closely match. The identifying means reads a signal at a plurality of different time points and generates a corresponding image of an associated space to produce a set of signal data and a corresponding set of image data; Means for fitting to a set of means, means for fitting a second curve to the corresponding set of image data, and means for comparing the first curve and the second curve. The device is at time t _sensor Means for reading the signal s from the real-time sensor, means for specifying the position of the signal s in the first curve, and specifying the point y in the second curve corresponding to the signal s in the first curve , Time t _sensor And means for estimating the position of the target site at. The apparatus also provides means for periodically taking a new X-ray image of the relevant space during treatment and simultaneously reading the corresponding signal, and updating the correlation with the new X-ray image and the corresponding signal. Means for doing so.

Yet another aspect of the invention is a method for locating an internal target site of a patient undergoing movement during treatment comprising:
Generating a plurality of first images of a patient's associated space prior to treatment, wherein the associated space includes the internal target site and each of the plurality of first images indicates the internal target site; ,
Generating a second image of the relevant space at a predetermined time during treatment;
Determining the position of the internal target site at the predetermined time by identifying one of the plurality of first images that most closely matches the second image. In the method, the patient motion includes a periodic respiratory motion characterized by a respiratory cycle, and the step of generating the plurality of first images includes a first computer tomogram taken at a first time point of the respiratory cycle. Generating a graphic scan; generating a second computer tomography scan taken at a second time point of the respiratory cycle; and at a third time point between the first time point and the second time point, Generating at least one computer-generated intermediate three-dimensional image representing the relevant space. The method further includes calculating a plurality of digitally reconstructed X-ray images for each of the plurality of first images, each of the plurality of digitally reconstructed X-ray images having an associated space when viewed from a specific viewing angle. Can be represented. Each of the first images may show a certain position of the target site at a corresponding point in the respiratory cycle, and each digitally reconstructed X-ray image may show the rotation and translation of the target site.

  The present invention is particularly applicable to devices and methods for directing a radiation beam toward an internal target site without implanting an internal fiducial, in which the present invention is described. However, the apparatus and method according to the present invention may be applied to other types of medical procedures using different types of medical devices, such as biopsy needle positioning, ablation, ultrasound or other focus energy therapy, or lasers. It will be appreciated that it has excellent utility for positioning a laser beam for beam therapy. Before describing the present invention, a general radiosurgical instrument will be described to provide a better understanding of the present invention.

  As used herein, a “target site” is an area to which treatment (eg, radiation) is to be directed. The target site is located within a “relevant space (meaning volume)” that means an internal region surrounding the target site, which can include bone and soft tissue around the target site.

  The method according to the invention comprises steps that can be carried out before the treatment (treatment) and steps that are carried out during the treatment. Prior to treatment, at least two series of CT scans are taken and calculated. Each CT scan corresponds to a specific point in the patient's respiratory cycle. Each series of CT scans may be an actual CT scan or a combination of an actual CT scan and a computer generated intermediate CT image. In order to generate an intermediate CT image, at least two CT scans are first acquired. A composite deformed image is then calculated from the actual CT scan. Each CT scan and computer-generated CT image shows the target site. This target site can be marked before treatment. A set of digitally reconstructed X-ray images (DRR) (a set of DRRs) is formed for each CT scan and / or computer-generated CT image. Each DRR indicates a target site from a specific angle.

  During treatment, live (raw) stereoscopic X-ray images are taken periodically, for example every 10 seconds. The target site will not be clearly visible in the x-ray image. However, the exact position of the target site is determined by comparing each live X-ray image with the above-mentioned DRR set, finding the DRR that most closely matches (matches / matches) the live X-ray image, and the CT in which the DRR was generated It can be determined by specifying a scan or CT image. A CT scan or CT image shows the target site and thus the location of the target site. Based on the viewing angle (viewing angle or viewing angle) associated with the most consistent DRR, the exact angle or translational shift of the patient at the time the live X-ray image was taken is also determined. Using DRR, both translational / rotational movements of the patient's body and the patient's current breathing state can be inferred from the live x-ray image. This treatment does not require any fiducial (reference) to be implanted and only requires an X-ray image during the procedure.

  Since X-ray images alone may be too slow to detect fast breathing, a non-fiducial (non-fiducial) target tracking method according to the present invention is real time to track the target site in real time. Can be combined with sensors. The real-time sensor may be an outer marker (marker) that is coupled to the patient's body part that moves simultaneously but not necessarily in the same direction or to the same extent as the target site moves. Because a real-time sensor is used, a correlation model is generated that associates the signal from the real-time sensor with the location of the target site, preferably prior to treatment. The correlation model takes an X-ray image of the target site and simultaneously reads the signal from the real-time sensor and then uses the X-ray image to identify the best matching DRR and associated CT scan or CT image Is generated by Since the real-time sensor is read at the same time as the X-ray image is taken, the CT scan or CT image specifies the position of the target site at the time the signal is read from the real-time sensor. After the set of data points is acquired, the location of the target site can be associated with signal readings from the real-time sensor. This pre-treatment correlation procedure is similar to the live no-fiducial targeting method performed during treatment in that the X-ray image is matched (matched) with the DRR to locate the target site. This correlation procedure differs from live target site tracking performed during treatment in that the X-ray image is taken and the signal from the real-time sensor is read at the same time. Once the correlation model is established, the location of the target site can be inferred from the real-time sensor signal. Because real-time sensor signals are easy to acquire and can be read more frequently than x-ray images are processed, the use of real-time sensors ensures that the target site is tracked almost continuously or in real time. Enable. Further details of the above correlation procedure are given in US Pat. No. 6,144,875. This document is incorporated herein in its entirety.

  FIG. 1 shows an example of a stereotaxic radiation therapy apparatus 10. The radiation therapy device 10 may include a data processing device 12 such as a microprocessor and a storage device 13 that may store image data and mathematical data associated with a target site within the patient 14. The image data can be captured into a data processing device before treatment or during surgery. The radiation therapy device 10 may also include a beam generating device 20. When activated, the beam generator 20 emits a parallel surgical ionization beam toward a target site within the patient 14. The parallel surgical ionization beam may be strong enough to necrotize the target site. Various other beam generators that generate ionizing radiation or heavy particle beams such as linear accelerators, preferably X-ray linear accelerators, can be used. Such X-ray beam generators are commercially available. The beam generator 20 can be operated by an operator using a switch (not shown) connected to the beam generator.

Radiation therapy apparatus 10 may also include a first diagnostic beam 26 and a second diagnostic beam 28 through <br/> the internal target portion position (passing) stereoscopic X-ray image apparatus for. These diagnostic beams are positioned at a predetermined (pre-set) non-zero angle relative to each other. These diagnostic beams can be generated by the first X-ray generator 32a and the second X-ray generator 32b, respectively. The image receiver 34 can receive the diagnostic beams 26, 28 and generate an image from the diagnostic beams. This diagnostic image can be supplied to the microprocessor 12 and compared with the three-dimensional image. In certain embodiments, two separate receivers can each receive one of the diagnostic beams 26 and 28.

  The radiation therapy device 10 may also include a device for adjusting the relative position of the beam generator 20 and the patient 14 such that the ionizing beam is continuously focused on the target site. In the particular radiotherapy device shown, the positioning of the beam generator relative to the patient can be altered by a processor-controllable robot arm mechanism 40 and / or a movable operating table having an inclined top surface 38. The robot arm mechanism allows the beam generator to move freely around the patient's body, including up and down, vertical and horizontal (left and right) directions along the patient.

  The radiation therapy device 10 may also include a real-time detection system for monitoring movements outside the patient 14. The real-time detection system includes one or more real-time sensors 42 coupled to a body part outside the patient 14 and a sensor reader 44 that periodically captures readings from the real-time sensor 42. Real-time sensor 42 readings indicate movement of the outer body part of patient 14. This real-time detection system can be any system that can be used to associate the real-time sensor 42 with a breathing pattern with a response time / reactivation (regeneration) time of less than 250 ms. Some commercially available sensors that can be used as the real-time sensor 42 include an infrared tracking system made by Northern Digital (Ontario, Canada), as well as force sensors, anemometers, strain gauges, laser distance sensors. And various sensors based on other physical principles such as tactile, auditory (acoustic) / ultrasonic, magnetic, mechanical or visual (optical) principles. Alternatively, the current state of breathing is detected by viewing video images of chest and / or abdominal movements of patient 14 or by detecting airflow or temperature that emulates patient 14's mouth and / or nose. Can be measured. The real-time detection system is coupled to the processor 12 so that the processor 12 can use the reading of the real-time sensor 42 to establish the correlation.

  FIG. 2 is a block diagram of the radiotherapy apparatus 10, which includes the microprocessor 12, the storage device 13 (for example, a tape drive mechanism), the beam generator 20, the robot arm 40, and the X-ray camera as described above. 30, 32, 34 and 36 and an operator control console 24. In addition, the device 10 may include a real-time sensor tracking system 50 that tracks the position of real-time sensors attached to the skin of the patient 14. The apparatus 10 may also include an operator display 48 for watching the treatment progress and managing the treatment. Further details of the radiosurgery device may be found in US Pat. No. 5,207,223 owned by the assignee of this application. This document is incorporated herein by reference.

  FIG. 3 shows a plot 70 of the patient's breathing pattern. The depiction portion of the breathing pattern includes two breathing cycles: cycle 72 and cycle 74. A single breathing cycle includes the full range of diaphragm and chest wall movements in the breathing pattern. Therefore, time point A to time point A is one respiration cycle as well as time point B to time point B. Because each respiratory cycle is substantially similar to each other, time point B in respiratory cycle 72 is related to an internal anatomy that is substantially similar to time point B in respiratory cycle 74. Therefore, as long as the patient does not move, the target site is at substantially the same position at each time point A and at substantially the same position at each time point B. Also, the target site is displaced from a position at time A to a position at time B in substantially the same manner in all cycles. Thus, it is possible to create a set of related space images in which there is an image representing the actual position of the related space at almost every point in the patient's breathing pattern. Such a breathing pattern represents the individual anatomy of each patient. For some patients training regular breathing, it may be necessary to maximize reproducibility.

  Because the patient's breathing pattern is substantially periodic, the processor 12 (see FIG. 1) can be programmed to issue specific commands at one or more preselected points in the breathing pattern. For example, the processor 12 can be programmed to take a CT scan at some point in the respiratory pattern. This programmed CT scan operation can include two general steps. That is, a step (first step) of determining (identifying) a patient's breathing pattern, and a step (second step) of programming the processor 12 to trigger the CT scanner at a certain point in the breathing pattern. In the first step, the patient's breathing pattern may be established by using any well-known method or instrument, such as the real-time detection system shown in FIG. A real-time sensor 42 (see FIG. 1) that emits a signal representative of movement of a body part (eg, chest wall) is coupled to the patient's body. After a critical number of signals are received, the signals are processed to reveal a breathing pattern.

  In the second step, a point in the pattern is selected based on the desired number of CT scans, and the processor 12 is programmed to trigger the CT scanner at the selected point. For example, a first CT scan can be taken at time A, a second CT scan can be taken at time B, and three CT scans can be taken at equal time intervals between time A and time B. This results in a total of five CT scans taken at different points in the respiratory cycle. More specifically, three CT scans between time points A and B can be taken at points C, D and E shown in FIG. Alternatively, the time point in the breathing pattern can be selected based on the position of the body part being tracked. In this case, the first CT scan is taken when the chest wall is at its highest level, the second CT scan is taken when the chest wall is a distance Ad below the highest level, and the third CT scan is taken at a distance from the highest level. It may be taken at a point 20d lower. Those skilled in the art will understand the following points. That is, this breathing pattern based triggering (triggering) method is not limited to being used with a CT scanner, and the processor 12 can issue commands to different devices, or the command set can be It can execute itself. For example, the processor 12 triggers the CT scanner at time points A and B and generates a composite image of the scanned related space for other time points between time points A and B of the respiratory cycle. obtain.

  This method, i.e., selecting a point in time of the breathing pattern and programming the processor 12 to trigger the CT scanner at this point in time, is not limited to use in a radiosurgical environment. For example, this method can be used to improve conventional CT scans that are known to be adversely affected in quality by patient motion (eg, respiratory motion). This method allows a person to obtain an improved 3D image without motion artifacts, eg, for diagnostic purposes. If a CT scan is taken using this method, it is clearly known at what point in the respiratory cycle the CT scan is taken.

  The internal position determination method according to the present invention includes a pre-processing procedure 100a and another pre-processing procedure 100b, one of which is preferably performed prior to radiation therapy (treatment). FIG. 4A and FIG. 4B each represent this pre-processing procedure 100a and another pre-processing procedure 100b. 5 and 6 schematically show exemplary three-dimensional images acquired during the pre-processing procedures 100a and 100b. FIG. 7 schematically illustrates a variation that can be used to generate several three-dimensional images in another pre-processing procedure 100b.

  FIG. 4A is a flowchart of the pre-processing procedure 100a, which is performed prior to processing for determining the location of the target site according to the present invention. The pre-processing procedure 100a begins when the patient is subjected to multiple CT scans to create a three-dimensional image showing the patient's associated space (ie, bone, soft tissue and target site) at different points in the patient's respiratory cycle (step 102). ). The CT scan (s), each indicating a target site, can be taken in the manner previously described in connection with FIG. The location of the target site is determined for each of those CT scans (step 110) and stored (step 112). In addition, a set of DRRs is generated for each CT scan. Each DRR represents how the associated space looks from a particular angle.

  More specifically, each CT scan taken in step 102 represents the patient's internal space (related space) at a particular point in the patient's respiratory cycle. For example, if exactly two CT scans are taken, the first CT scan can be taken at time A of the respiratory cycle and the second CT scan can be taken at time B. If the relevant space is near the lungs or liver, time points A and B can be, for example, the maximum and minimum inhalation points of the patient's respiratory cycle. The location of the target site is identified in each of these CT scans, for example as coordinates in the treatment room (step 110). Then, the specified position is stored in the storage device 13 (see FIG. 1) of the radiation therapy apparatus 10 (step 112). The person skilled in the art will understand the following points. That is, although the preprocessing procedure 100 has been described as including a CT scan, the CT scan can be taken with other three-dimensional images, such as magnetic resonance (MR) images or ultrasound images, showing bones and tissues in the relevant space. Can be replaced.

  Because patients may shift their body during treatment, it is important to obtain views of the target site from different angles in order to increase the accuracy of locating the target site. Each CT scan is used to generate a set of digitally reconstructed X-ray images (DRR) to take into account any movement of the patient under treatment (step 108). Each DRR is an image obtained by calculating a two-dimensional projection through a three-dimensional image. Therefore, DRR is a composite image obtained by calculation. Two-dimensional projection through a three-dimensional CT scan is similar to the physical process of taking an X-ray image. Therefore, DRR looks similar to an X-ray image. In fact, if the DRR is obtained from exactly the same angle as the corresponding x-ray image, the similarity will be quite close. In this embodiment, each DRR is a composite two-dimensional image showing how the three-dimensional image prepared in steps 102 and 104 looks from a specific angle.

  A set of DRRs, all calculated from different angles of a single 3D image, is similar to a set of X-ray images taken from these angles. Thus, preferably, these DRRs indicate target sites from a set of angles at which the X-ray generators 32a, 32b (see FIG. 1) view (capture) the target sites. Since the patient may move spontaneously or unconsciously during treatment, the DRR needs to show how the relevant space looks from different angles. Preferably, there is a sufficient (number) of DRRs in a set so that there are DRRs corresponding to almost every location that the patient 14 can move to during treatment. Also, the DRR set may include as many DRRs as a person skilled in the art will consider sufficient.

  FIG. 4B shows another pre-processing procedure 100b according to the present invention. This other preprocessing procedure 100b is similar to the preprocessing procedure 100a of FIG. 4A, except that the 3D image is a combination of an actual CT scan and a computer generated intermediate 3D image. In step 102, multiple (eg, two) CT scans of the relevant space are taken at time points A and B of the respiratory cycle. These CT scans are then used to calculate a series of intermediate 3D images by calculating a composite deformed image from the actual CT scan taken during step 102 (step 104). Each intermediate three-dimensional image shows the position of the target site in the respiratory state between time A and time B. The calculations for generating these intermediate three-dimensional images can be performed off-line by any well-known method such as thin plate spline, warping, interpolation or extrapolation. The location of the target site is marked in each of these CT scans and intermediate 3D images (step 106). This marking can also be done off-line. Both the CT scan taken in step 102 and the intermediate 3D image obtained in step 104 are referred to herein as “3D images”.

  In one exemplary embodiment, two CT scans are acquired at step 102 and ten intermediate 3D images are generated at step 104. Next, forty DRRs are generated in step 108 for each of these three-dimensional images. Since there are a total of 12 3D images, this means that a total of 480 DRRs have been generated. In other words, each three-dimensional image showing a certain point in the respiratory cycle is seen (captured) from 40 different angles. As will be described later with reference to FIG. 8, DRR is used to match (match or match) a live X-ray image with a three-dimensional image, which is then used to determine the position of the target site. Used to determine.

  FIG. 5 schematically shows a CT scan 210a taken at time point A of the respiratory cycle in stage 102 of the preprocessing procedure 100. FIG. FIG. 6 shows a CT scan 210b, which was also taken at point B of the respiratory cycle in stage 102 of the preprocessing procedure 100. In particular, FIGS. 5 and 6 show cross-sectional views of the associated space 200, the soft tissues 202a and 202b, and the target site 204 located between the soft tissues 202a and 202b. The soft tissues 202a and 202b are aligned in the X direction according to the coordinate system 214. At time A of the respiratory cycle, the target site 204 is closer to the soft tissue 202a than to the soft tissue 202b, as shown in FIG. However, at point B of the respiratory cycle, the target site 204 is closer to the soft tissue 202b than the soft tissue 202a, as shown in FIG. Thus, the target site 204 is displaced along the X direction between time points A and B of the respiratory cycle. Moreover, the shape of the related space 200 is different between FIG. 5 and FIG. More specifically, the associated space 200 and the soft tissues 202a and 202b contract along the Y direction when traveling from time A (shown in FIG. 5) to time B (shown in FIG. 6). Accordingly, as the respiratory cycle progresses from time point A to time point B, the associated space 200 shrinks along the Y direction, and the target site 204 is displaced along the X direction.

  FIG. 7 schematically illustrates the generation of an intermediate 3D image in stage 104 of FIG. 4B. Two exemplary intermediate 3D images 220 and 222 are generated based on the two CT scans 210a and 210b. Three-dimensional images 220 and 222 are formed by continuously transforming a CT scan 210a taken at time A of the respiratory cycle into a CT scan 210b taken at time B. Therefore, the three-dimensional images 220 and 222 represent an intermediate stage through which the associated space 200 and target site 204 pass during the transition from time A to time B in the respiratory cycle. As described above in connection with FIG. 3, for purposes of illustration herein, time point A is the time of maximum inhalation and time point B is the time of maximum exhalation. Therefore, as the patient 14 exhales, the associated space 200 gradually decreases along the Y direction, and the target site 204 gradually approaches the soft tissue 202b.

  FIG. 8 is a flowchart illustrating a target tracking procedure 130 without fiducials. This target tracking procedure 130 allows the location of the target site to be determined according to the present invention without embedding fiducials. During treatment, live stereoscopic X-ray images are taken periodically at time intervals Δt (step 131). Since the x-ray image itself does not represent a target site, live x-rays must be associated with an appropriate three-dimensional image in order for the target site to be located. Each of these X-ray images is compared with the DRR prepared during the preprocessing procedures 100a, 100b (step 132). Since X-ray images do not represent everything in the associated space that the 3D image represents (eg, the target site), what is matched (matched) is an association that appears in both the X-ray image and the DRR It is the position of a part of the space. This comparison may be performed by the processor 12 for each live x-ray image using well known image comparison techniques. The image comparison techniques include, but are not limited to, mutual information, cross-correlation and image subtraction. With one of these techniques, each DRR and live X-ray image is compared pixel by pixel. This comparison may involve subtracting the gray level pixel values of both images for each pixel location. The cumulative difference in gray level gives an error signal that characterizes the distance between the DRR and the X-ray image. Those skilled in the art will understand how to implement an appropriate comparison technique.

  Through this comparison, the DRR that most closely matches the X-ray image is selected (step 134). Since all DRRs are associated with a 3D image, the associated 3D image is identified (step 136). In addition, the appropriate angle associated with the most consistent DRR must be identified (step 138). Based on the identified three-dimensional image and viewing angle, the position of the target site is determined (step 140). The viewing angle is then added to this target location (also step 140). Once the position of the target site as seen from the angle of the X-ray imaging device is known, the position of the target site can be accurately determined. Next, the location of the target site is estimated and determined relative to the treatment room coordinates (step 128). Since the breathing pattern is substantially periodic, the location of the target site can be predicted even after acquisition of a critical number of data points.

  The non-fiducial procedure 130 of FIG. 8 provides the significant advantage of locating the target site without an implant fiducial, but X-ray images alone are too slow for fast breathing. Because of the wax, the procedure 130 does not allow real-time tracking. Too frequent X-ray imaging (X-ray imaging) may expose the patient to excessive radiation, so the time interval Δt at which X-ray images are taken can be on the order of 10 seconds. Target site positioning every 10 seconds does not give accurate beam guidance. This is because the target site can move out of the beam radius (range) within 10 seconds. The treatment beam needs to be adjusted between the X-ray images with a time interval shorter than Δt. Real-time sensors that can be easily tracked can be implemented to achieve tracking of closer target sites. The real-time sensor provides real-time measurement data, ie, measurement data with negligible time lag between the motion of the patient's body part and the “report” of the motion. Determination of the location of the target site based on the location of the real-time sensor allows real-time determination of the location of the target site. The use of real-time sensors is less invasive, less expensive and less complex overall than the use of internal fiducials, so tumors can be based on real-time sensors without using internal fiducials. A function to determine the position is desired.

  To overcome the slowness of X-ray imaging that makes real-time tracking difficult, real-time sensors can be used with a non-fiducial target tracking procedure 130. In order for a real time sensor to be used with a non-fiducial target tracking procedure 130 to locate a target site in real time, a correlation needs to be established between the real time sensor and the location of the target site.

FIG. 9 is a flowchart illustrating a sensor-target site correlation procedure 120, which is a procedure for establishing a correlation model between real-time sensor readings and target site locations. In the treatment room, the patient is placed on the tiltable top surface 38 (see FIG. 1) and within the field of view of at least two x-ray generators 32a, 32b (see FIG. 1). In real-time target tracking, a real-time sensor can be coupled to the patient's outer body part (eg, skin), that is, activated after the patient 14 is placed in the treatment room (stage 122). As described above in connection with FIG. 1, the real-time sensor may be any sensor that correlates to respiratory action with a response time or reactivation time of less than 250 ms. Moreover, the real-time sensor should emit a new signal at least ten times per second. The signal from the real-time sensor can be read at a time interval Δt _sensor that is shorter than the time interval Δt. At that time, a stereoscopic X-ray image is taken (step 124). At the same time an X-ray image is taken, the signal from the real-time sensor is read, and a time stamp can be recorded in the reading. Stage 124 is repeated at time interval Δt. The sensor reading interval Δt _sensor need not be completely constant as long as each successive sensor reading is obtained sufficiently dense in time (ie, as long as Δt _sensor is sufficiently short). The same is true for x-ray imaging.

  The stereoscopic X-ray image is then compared to the DRR obtained during the preprocessing procedure 100a, 100b (step 126). This comparison identifies the best matching DRR, which points to the 3D image from which this DRR is generated. Since the position of the 3D image is marked during the preprocessing procedure, the position of the target site is determined from the 3D image (step 128). Using the determined target site location, the data points collected in step 124 are converted into data points with the target site location and corresponding real-time sensor readings to generate a point cloud. The processor 12 (see FIG. 1) in the radiosurgery device can fit the first curve to the point generated by the real-time sensor and fit the second curve to the point generated for the target location. These first and second curves allow real-time sensor readings and target locations to be correlated with each other (step 130), creating a correlation model that is ultimately used to track the target site during treatment.

  Another way to correlate target location with real-time sensors is to perform interpolation or other known mathematical interpolation methods to establish correspondence between two sets of data after calculating the point cloud. Using a neural network.

FIG. 10 shows a real-time tracking procedure 150 for real-time tracking of the target site being treated. The real-time tracking procedure 150 begins by acquiring a reading from the real-time sensor at times t _sensor1 , t _sensor2 , t _sensor3 ,... _Divided by time intervals that are not necessarily constant (step 152). For ease of explanation, the time interval is expressed herein as Δt _sensor . Here, Δt _sensor is a certain time range rather than an accurate and constant value. Since the real-time sensor signal can be read more frequently than the interval at which X-ray images are taken (Δt> Δt _sensor ), real-time sensor information can be acquired between successive X-ray images. For example, Δt _sensor can be 50 ms and Δt can be 10 seconds.

The system reads the signal s from the real-time sensor at time t _sensor1 . The position of the target site at time t _sensor1 is then obtained based on this real-time sensor reading (step 156). This real-time sensor reading is based on the correlation established between the real-time sensor reading and the target site location established during the sensor-target site correlation procedure 120. More particularly, the signal s is fitted to a first curve of pre-generated real-time sensor readings. Next, the position y of the target site corresponding to the sensor signal s is determined by identifying a point on the second curve corresponding to the position of s on the first curve, or by a known interpolation method. If there are multiple real-time sensors, this process can be performed for each real-time sensor. Thus, the location of each real-time sensor obtained in step 152 is matched to one of the real-time sensor readings in the correlation model, and the location of the target site is estimated from the real-time sensor reading.

  As described above, the real-time sensor signal is read frequently (for example, every 50 ms). Based on the sensor-target position correlation, the position of the target site to be determined is read as often as the sensor signal. Therefore, in the present invention, it is not necessary to actually image the internal target site on a real-time basis in order to track the target site almost continuously.

Additionally, when the signal s is read from the real time sensor, a live X-ray image can be taken at time t ₀ (stage 154). Preferably, the X-ray time interval Δt is a multiple of the sensor reading time interval Δt _sensor so that X-ray imaging and sensor reading occur simultaneously after a certain number of sensor readings. This new X-ray image, which is an additional data point, can be added to the point cloud and can be used to improve or update the correlation model. This constant update prevents any change in the breathing pattern during treatment from compromising the accuracy of the treatment.

It should be noted that the present invention is not limited to a specific number of sensors and sensors that track respiratory motion. The number of sensors to be used can be determined based on the degree of accuracy desired, ie accuracy. Multiple sensors can lead to multiple levels of correlation, increasing tracking reliability and accuracy. In one embodiment, three sensors can be used. That is, a first sensor read at the time interval Δt _sensor , a second sensor read at the time interval Δt, and a third sensor read at another time interval between Δt _sensor and Δt. Additional sensors can follow the patient's movement caused by something other than the respiratory action that also moves the target site. For example, aortic pulsations or cardiac pulsation cycles may be tracked to account for target site movement due to cardiac motion. If the movement tracked by the additional sensor is periodic, this can be dealt with in a manner similar to the manner in which breathing movement is handled. If desired, the sensor can be used to track only movements other than respiratory movements.

  The present invention allows real-time tracking of the target site being treated based on the location of the real-time sensor. Moreover, the present invention allows the target site to be tracked during treatment based solely on x-ray images. The X-ray image is acquired when the position of the real-time sensor is determined. In addition, each DRR is associated with an intermediate 3D image prior to treatment. Therefore, only the tasks that the processor 12 needs to perform during treatment find the DRR that is most consistent with the X-ray just taken. The simplicity (simpleness) of the task and the short processing time depend on when the processor 12 determines the position of the target site and when the beam generator 20 (see FIG. 1) Minimizes the time lag from the point of adjustment. In the meantime, marker tracking is performed at a high frequency so that the treatment beam can be adjusted almost continuously to ensure that the treatment beam is accurately directed to the target site for the entire duration of the treatment. Because marker tracking is non-invasive and inexpensive, the present invention allows for a non-invasive and inexpensive alternative that increases the accuracy of radiation therapy.

  Although the foregoing description relates to a particular embodiment of the present invention, those skilled in the art will appreciate that the embodiment can be modified without departing from the scope of the principles and spirit of the present invention. The scope of the invention is defined by the claims.

1 shows an example of a radiosurgery treatment system that can be used with the present invention. It is a block diagram showing the treatment system of FIG. Fig. 3 shows a breathing pattern including multiple breathing cycles. 4 is a flowchart of a pre-processing procedure executed before real-time tracking of a target site according to the present invention. 4 is a flowchart of another pre-processing procedure performed prior to real-time tracking of a target site according to the present invention. 1 schematically shows the location of a target site at time A of the respiratory cycle. 3 schematically shows the position of the target site at time B of the respiratory cycle. 3 schematically illustrates the formation of an intermediate 3D image by deformation from a first 3D image to a second 3D image. FIG. 6 is a flow chart depicting a fiducial-free target tracking procedure for determining a target location during treatment without using an implantable fiducial. FIG. 5 is a flow chart representing a correlation method for associating real-time sensor readings with target locations. FIG. 6 is a flowchart representing a real-time tracking procedure for real-time tracking of a target site during treatment. FIG.

Explanation of symbols

10 Stereotaxic radiotherapy equipment 12 Data processing equipment (microprocessor, etc.)
DESCRIPTION OF SYMBOLS 13 Memory | storage device 14 Patient 20 Beam generator 24 Operator control console 26 1st diagnostic beam 28 2nd diagnostic beam 30 Internal target site | part 32a 1st X-ray generator 32b 2nd X-ray generator 34 Image receiver 38 Upper surface 40 Robot arm mechanism 42 Real-time sensor 200 Related space 202a, 202b Soft tissue 204 Target site 210a, 210b CT scan 220, 230 Intermediate 3D image

Claims (4)

A method for locating an internal target site during treatment comprising:
Generating a plurality of first images of a patient's associated space, the associated space including the internal target site, and each of the plurality of first images indicating an internal target site;
Generating a live image of the relevant space at a predetermined time during treatment ;
Determining the position of the internal target site at the predetermined time by identifying one of the plurality of first images that most closely matches a live image. A device for tracking the position of an internal target site within a patient,
The live image of the relevant space including the internal target site and configured imaging device to generate,
A storage device for storing the 3-dimensional image showing a plurality of three-dimensional image in which the image is a by respective three-dimensional image is generated in advance of the associated spatial及beauty internal target sites that were generated prior to treatment,
A processor coupled to the image device and the storage device,
A processor includes a first set of computer readable instructions for selecting an optimal 3D image that most closely approximates the live image, and an internal target site at the time the live image is generated based on the optimal 3D image And a second set of computer readable instructions for determining the position of the computer. A device for tracking the location of a target site inside a patient,
Means for generating a plurality of first images of a relevant space inside a patient, wherein the relevant space includes a target site, and each of the plurality of first images indicates a target site;
Means for generating a live image of the relevant space at a predetermined time during treatment ;
Means for determining a position of an internal target site at the predetermined time by identifying one of the plurality of first images that most closely matches a live image. A method for locating an internal target site of a patient undergoing movement during treatment comprising:
Generating a plurality of first images of a patient's associated space prior to treatment, wherein the associated space includes the internal target site and each of the plurality of first images indicates the internal target site; ,
Generating a second image of the relevant space at a predetermined time during treatment;
Determining the position of the internal target site at the predetermined time by identifying one of the plurality of first images that most closely matches the second image.